Bug: Duplicate Responses in 'generate' mode

[schmichri]

In generate mode NLWeb returns the same answer twice.

Bug Report: Duplicate Responses from Ask Endpoint

Issue Description

When asking a question through the /ask endpoint, the system returns the same response twice. The response data with format {"message_type": "nlws", "answer": "something"} is being sent twice for a single request.

Steps to Reproduce

1. Setup NLWeb + Import Data
2. set the generate mode in index.html chat_interface = new ChatInterface('', 'nlwebsearch', 'generate');
3. Send a request to the /ask endpoint
4. Observe the response stream

Expected Behavior

The system should send a single response for each request, containing the answer and any relevant item descriptions.

Actual Behavior

The system sends two identical responses:

1. An initial response with just the answer
2. A second response with the same answer and item descriptions

Root Cause

In the synthesizeAnswer method of the GenerateAnswer class, the code was sending an initial message with just the answer before processing URLs:

# Create initial message with just the answer
message = {"message_type": "nlws", "answer": answer, "items": json_results}
logger.info("Sending initial answer")
await self.send_message(message)

Then, after processing URLs and item descriptions, it would send another message with the same answer plus the item descriptions.

Solution

The fix involved removing the initial message sending, ensuring that the response is sent exactly once in all cases:

1. If there are no ranked answers, it sends a message with a default answer and empty items list
2. If there are URLs and description tasks, it sends a message with the answer and item descriptions
3. If there are no URLs found in the synthesis response, it sends a message with just the answer and empty items list
4. If there's an error, it sends an error message

This ensures that each request receives exactly one response, eliminating the duplicate message issue.

[rvguha]

We do want it to send the item as soon as possible, for reduced latency. Let me look into why the items are being sent again. The description exists right at the beginning for each item.

[johannespn]

I'm also getting duplicate entries on the default 'list' mode

: keep-alive

HTTP/1.1 200
Content-Type: text/event-stream
Cache-Control: no-cache
Connection: keep-alive
Access-Control-Allow-Origin: *
Access-Control-Allow-Methods: GET, OPTIONS
Access-Control-Allow-Headers: Content-Type

data: {"message_type": "asking_sites", "message": "Asking xxx-enriched", "query_id": "query_1750155450131_960"}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]}

data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]}

data: {"message_type": "complete"}

while in the generated .jsonld, there is only one of each product.

[tim13337]

I think this is from the fast mode. that it is so fast that it gets caught up by the other thread - i think it would be good to generate more flow diagrams that everyone can easier understand how it works: #258

[rvguha]

I believe this problem has been fixed. Let me know if you are still seeing it.

…

On Tue, Jun 17, 2025 at 3:23 AM Johannes ***@***.***> wrote: *johannespn* left a comment (microsoft/NLWeb#190) <#190 (comment)> I'm also getting duplicate entries on the default 'list' mode : keep-alive HTTP/1.1 200 Content-Type: text/event-stream Cache-Control: no-cache Connection: keep-alive Access-Control-Allow-Origin: * Access-Control-Allow-Methods: GET, OPTIONS Access-Control-Allow-Headers: Content-Type data: {"message_type": "asking_sites", "message": "Asking xxx-enriched", "query_id": "query_1750155450131_960"} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-grape", "name": "High Waist Leggings | Grape", ... }]} data: {"message_type": "result_batch", "results": [{"url": "https://xxx.com/products/high-waist-leggings-azure", "name": "High Waist Leggings | Azure", ... }]} data: {"message_type": "complete"} while in the generated .jsonld, there is only one of each products. — Reply to this email directly, view it on GitHub <#190 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABICKCUZWMXHS6HXFBG6DB33D7UCBAVCNFSM6AAAAAB67SEXYCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDSNZZHAYDSMRVGA> . You are receiving this because you were assigned.Message ID: ***@***.***>